UT theoretical chemistry code forum

Hi,

I've been using the CI-NEB framework in VASP to compute the MEP for Au atoms diffusing on the TiO2 rutile (110) surface. Atoms below a couple of layers below the surface are frozen at their bulk positions. I would now like to use the dynamical matrix code to compute the vibrational modes for the transition state and minima. However, it would be very computationally expensive (and likely unnecessary) to include displacements of all atoms in the unfrozen atomic layers in the DISPLACECAR. I was wondering how has been done for similar systems with surfaces (for example by the Henkelman group). Is there a rule of thumb for how large a radius is generally used when creating the DISPLACECAR with dymselsph.pl?

Thanks!

If you are interested in calculating a prefactor, I would suggest using the dymseldsp.pl script. This selects atoms which have moved the most in the transition. This is typically much better than a spherical radius because it selects the atom which are involved in the transition and contribute to the prefactors. You can also use some intuition for how large a displacement is important. 0.5 Ang is clearly significant whereas 0.01 is probably not.